speechllm-1.5B / config.py
shangeth's picture
Upload model
898044c verified
raw
history blame
509 Bytes
from transformers import PretrainedConfig
class SpeechLLMModelConfig(PretrainedConfig):
model_type = "custom_model"
def __init__(self, **kwargs):
super().__init__(**kwargs)
self.audio_enc_dim = 1024
self.llm_dim = 2048
self.audio_processor_name = "microsoft/wavlm-base"
self.audio_encoder_name = 'microsoft/wavlm-large'
self.llm_model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
self.llm_model_checkpoint = "hf_repo/llm_model_checkpoint"